There seems to be some issues with Anaconda2 and OpenCV3, especially with the ffmpeg bindings. As they are important for me and I couldn't get them running with anaconda's python, everything is made for the preinstalled python.

System update

sudo apt-get update
sudo apt-get upgrade

Install NVIDIA Driver (GTX 970)

sudo apt-get dist-upgrade
sudo apt-get install build-essential
sudo apt-get install linux-source
sudo apt-get install linux-headers-generic

Open /etc/default/grub and change the line of GRUB_CMDLINE_LINUX_DEFAULT to GRUB_CMDLINE_LINUX_DEFAULT="nouveau.blacklist=1 quiet splash nomodeset"

sudo update-grub2
sudo apt-get remove nvidia* #incase of not comming from a fresh installation
sudo apt-get autoremove #ensures no former installation clashes with new install
sudo reboot

After the restart download the newest nvidia-driver from the nvidia homepage. Right-click on the downloaded file and change it to executable.

Then open /etc/modprobe.d/blacklist.conf and add the following lines to the end of the file:

blacklist vga16fb
blacklist nouveau
blacklist rivafb
blacklist nvidiafb
blacklist rivatv
blacklist lbm-nouveu
options nouveau modeset=0
alias nouveau off
alias lbm-nouveau off

Then open the putty-terminal with Ctrl + Alt+ F1 and enter the following commands

sudo service lightdm stop #stops graphic session to enable nvidiainstallation
cd Downloads
sudo ./{the downloadedfilename.run} #follow the installation instructions (yes to all)
sudo nvidia-xconfig #(if you did not chose “yes” to this in the installation)

sudo nano /etc/default/grub #Change "GRUB_CMBLINE_LINUX_DEFAULT"-line to GRUB_CMDLINE_LINUX_DEFAULT="quiet splash nomodeset”
sudo update-grub2
sudo reboot

Check installation in the terminal by entering nvidia-smi or lshw -c video

Install CUDA

at the moment of writing this, the newest version of CUDA is 8.0

Download CUDA .deb file here

Preinstallation checks

# check if gcc is installed
gcc --version

#install kernel headers and dev packages
sudo apt-get install linux-headers-$(uname -r) #

then:

sudo dpkg -i cuda-repo-ubuntu1404-8-0-local_8.0.44-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda

Post-installation actions (added to ~/.bashrc)

export PATH=/usr/local/cuda-8.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64\${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

#reload bashrc config
source ~/.bashrc

#check installation
nvcc --version

Note: If error's occur regarding nvidia driver version, check if multiple drivers are installed. If yes, check this github issue

Install CUDNN

at the moment of writing this, the newest version of CUDNN is 5.1

Download CUDNN v5.1 Library for Linux here
Extract .tgz to folder of choice
Open Terminal and 'cd' to this folder

sudo cp -P include/cudnn.h /usr/include
sudo cp -P lib64/libcudnn* /usr/lib/x86_64-linux-gnu/
sudo chmod a+r /usr/lib/x86_64-linux-gnu/libcudnn*

Install git

sudo apt-get install git

If later on error's are encountered with runit and git-daemon-run do:

sudo apt-get purge runit
sudo apt-get purge git-all
sudo apt-get purge git
sudo apt-get autoremove
sudo apt-get update
sudo apt-get install git

Installing ffmpeg

In 16.04 it's simple again, installing via

sudo apt-get install ffmpeg

Installing OpenCV3

Actual version is 3.2.0, installation follows the guide of pyimagesearch but is copied for backup reasons and modified a bit (I install opencv globally, as I need it in nearly all of my projects).

Prerequisites:

sudo apt-get install build-essential cmake pkg-config
sudo apt-get install libjpeg8-dev libtiff5-dev libjasper-dev libpng12-dev
sudo apt-get install libavcodec-dev libavformat-dev libswscale-dev libv4l-dev
sudo apt-get install libxvidcore-dev libx264-dev
sudo apt-get install libgtk2.0-dev
sudo apt-get install libatlas-base-dev gfortran
sudo apt-get install python2.7-dev python3-dev

If not with Miniconda/Anaconda

sudo apt-get install python-pip
sudo pip install --upgrade pip
sudo pip install numpy

Download OpenCV 3.2.0 and Contrib-Package:

wget -O opencv.zip https://github.com/Itseez/opencv/archive/3.2.0.zip
unzip opencv.zip
wget -O opencv_contrib.zip https://github.com/Itseez/opencv_contrib/archive/3.2.0.zip
unzip opencv_contrib.zip

Build/Install OpenCV

cd /opencv-3.2.0/
mkdir build
cd build

For Python2.7 without Anaconda/Miniconda this works:

cmake -D CMAKE_BUILD_TYPE=RELEASE \
      -D CMAKE_INSTALL_PREFIX=/usr/local \
      -D INSTALL_PYTHON_EXAMPLES=ON \
      -D INSTALL_C_EXAMPLES=OFF \
      -D OPENCV_EXTRA_MODULES_PATH=~/opencv_contrib-3.2.0/modules \
      -D BUILD_EXAMPLES=ON ..

With Miniconda and for python3

cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=/home/USER/opencv-3.2.0 \
-D INSTALL_C_EXAMPLES=OFF \
-D INSTALL_PYTHON_EXAMPLES=OFF \
-D OPENCV_EXTRA_MODULES_PATH=/home/USER/opencv_contrib-3.2.0/modules \
-D BUILD_EXAMPLES=OFF \
-D BUILD_opencv_python2=OFF \
-D WITH_FFMPEG=1 \
-D WITH_CUDA=0 \
-D PYTHON3_EXECUTABLE=/home/USER/miniconda3/bin/python \
-D PYTHON_INCLUDE_DIR=/home/USER/miniconda3/include/python3.5m \
-D PYTHON_INCLUDE_DIR2=/home/USER/miniconda3/include/python3.5m \
-D PYTHON_LIBRARY=/home/USER/miniconda3/lib/libpython3.5m.so \
-D PYTHON3_PACKAGES_PATH=/home/USER/miniconda3/lib/python3.5 \
-D PYTHON3_NUMPY_INCLUDE_DIRS=/home/USER/miniconda3/lib/python3.5/site-packages/numpy/core/include ..

Then:

make -j8
make install

Maybe the cv2.so file is not correctly copied. With Miniconda3 I found "cv2.cpython-35m-x86_64-linux-gnu.so" inside opencv-3.2.0/build/lib/python3. Rename this to cv2.so and copie to miniconda/lib/python3.5/site-packages worked for me!

Then test correct installation and Video support by

cd
python

#inside python
import cv2
cv2.__version__ #should print '3.2.0'

vid = cv2.VideoCapture('test.avi') #load some test video
ret, frame = vid.read()
ret #should output True inside the console
cv2.imwrite('test.png', frame) #should save the frame as image

Install Caffe (Berkely's DL library)

Install dependencies see here

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler
sudo apt-get install --no-install-recommends libboost-all-dev
sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev

Clone Caffe's git repo and install python requirements

git clone https://github.com/BVLC/caffe.git
cd caffe
cat python/requirements.txt | xargs -L 1 sudo -H pip install
cp Makefile.config.example Makefile.config

Edit the Makefile.config and change following points:

uncomment CPU_ONLY := 1 (on machines w/o support graphic card)
uncomment OPENCV_VERSION := 3
Check Python Paths
Change INCLUDE_DIRS and LIBRARY_DIRS under #Whatever else your find.... to
- INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/
- LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial/
Change PYTHON_INCLUDE to correctly point to numpy/arrayobject.h (could be in /usr/local/lib/.. insted of /usr/lib/...)

Install/Build Caffe

make pycaffe -j8
make all -j8
make test -j8

Run tests to see if everything is correctly installed

./data/mnist/get_mnist.sh
./examples/mnist/create_mnist.sh
./examples/mnist/train_lenet.sh

If CPU_ONLY installation change /examples/mnist/lenet_solver.prototxt to: solver_mode = CPU

To work as a python importable package, we have to adapt the PYTHON_PATH variable -export PYTHONPATH=/path/to/caffe/python:$PYTHONPATH <-- to /.bashrc

If there is a "UserWarning: Matplotlib is building the font cache using fc-list" on the first import of caffe in python, do

rm -rf ~/.cache/fontconfig/   #worked for me

See this issue

Spyder IDE

Install spyder 2.3.9 with all optional dependencies. I had troubles using spyder 3, so this is how to setup spyders latest 2.3.9 version.

sudo apt-get install python-qt4 python-tk #required
sudo -H pip install ipython jedi==0.8.1 matplotlib pandas pep8 psutil pyflakes pygments
sudo -H pip install pylint qtconsole rope sphinx sympy zmq
sudo pip install spyder==2.3.9 #having troubles with spyder3 (debug object inspection)

If spyder, started from Launcher isn't able to import caffe add:

path/to/caffe/python to spyders path manager and restart spyder

Tensorflow, Theano and Keras

For tensorflow with only cpu-support, do the following:

sudo -H pip install tensorflow keras #keras will install theano as dependencie

For tensorflow with gpu support, we need cuda 8.0 with cudnn v5.1. We then can install tensorflow with gpu-support by the following piece of

sudo -H pip install tensorflow-gpu

Chrome Remote Desktop

Download and install Chrome Browser .deb here
Get Chrome Desktop Remote App here
Download & Install host components here
Create a file called ".chrome-remote-desktop-session" in the home folder and save with the following content

DESKTOP_SESSION=ubuntu XDG_CURRENT_DESKTOP=Unity XDG_RUNTIME_DIR=/run/user/$(id -u) exec /usr/sbin/lightdm-session 'gnome-session --session=ubuntu'.

Allow remote connection in the App

If on the first try of accessing the ubuntu machine via chrome remote desktop only the blank wallpaper shows up add following line to ~/.profile

export CHROME_REMOTE_DESKTOP_DEFAULT_SIZES=1920x1200 #or any resolution of choice

If session is already running, restart from terminal by

sudo /etc/init.d/chrome-remote-desktop stop
sudo /etc/init.d/chrome-remote-desktop start

Thunar File Manager

sudo apt-get install thunar

If Thunar is slow on start-up open Terminal and enter

sudo sed -i 's/AutoMount=true/AutoMount=false/' /usr/share/gvfs/mounts/network.mount

Jekyll

gpg --keyserver hkp://keys.gnupg.net --recv-keys 409B6B1796C275462A1703113804BB82D39DC0E3
\curl -sSL https://get.rvm.io | bash -s stable

#restart terminal
rvm get head 
rvm install ruby 
rvm list 

#restart terminal as login shell (edit -> profile preferences -> command -> run as login shell)
rvm use < ruby-version > 
gem install jekyll

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Ubuntu_16_04LTS.md

Ubuntu_16_04LTS.md

System update

Install NVIDIA Driver (GTX 970)

Install CUDA

Install CUDNN

Install git

Installing ffmpeg

Installing OpenCV3

Install Caffe (Berkely's DL library)

Spyder IDE

Tensorflow, Theano and Keras

Chrome Remote Desktop

Thunar File Manager

Jekyll

Files

Ubuntu_16_04LTS.md

Latest commit

History

Ubuntu_16_04LTS.md

File metadata and controls

System update

Install NVIDIA Driver (GTX 970)

Install CUDA

Install CUDNN

Install git

Installing ffmpeg

Installing OpenCV3

Install Caffe (Berkely's DL library)

Spyder IDE

Tensorflow, Theano and Keras

Chrome Remote Desktop

Thunar File Manager

Jekyll